Anima version is available!
Nova Furry XL
A 2d/2.5d furry checkpoint model that can have great details on any type of furs, scales and feathers, which aims to be XL-fied Peaki Furry
Rules
You cannot use the generated images for commercial use if it's not edited (or just turning it to black and white)
Advertising is always welcome
Recommended settings:
Sampler: Euler a for best performance
Steps: 20-50 steps
(Pony)
CFG Scale: 6-7
Prompt: score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_furry, BREAK, furry, anthro, detailed face eyes and fur, {Animal Type}
Negative Prompt: blurry, jpeg artifacts, username, watermark, signature, normal quality, worst quality, large head, low quality, text, error, missing fingers, extra digits, fewer digits, bad eye
(Illustrious)
CFG Scale: 3-5
Prompt: masterpiece, best quality, amazing quality, very aesthetic, high resolution, ultra-detailed, absurdres, newest, scenery, furry, anthro, {Prompt}, BREAK, depth of field, detailed fluffy fur, volumetric lighting
Negative Prompt: human, multiple tails, modern, recent, old, oldest, graphic, cartoon, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured, long body, lowres, bad anatomy, bad hands, missing fingers, extra digits, fewer digits, cropped, very displeasing, (worst quality, bad quality:1.2), bad anatomy, sketch, jpeg artifacts, signature, watermark, username, simple background, conjoined, bad ai-generated
Description
Changes
Extreme posing
Non-sepia filter
Reduced some blurriness
NoobAI EPS v1.1 + Illustrious v2.0-stable DARE applied
FAQ
Comments (47)
Love all your checkpoints so far. So what would constitute "extreme posing"?
Is this "furry" thing actually used often? 200k downloads is just mind boggling.
Yes, quite often, and it's not the only one. ^^
This world is doomed.
degenerates judging degenerates is crazy though
@nexusimperius I aint judging, just surprised lol. I don't really see the appeal.
@LuminalArt Well, to each their own. ;)
It's a damn good base model for a variety of applications, it follows style tags, it works well with style LoRAs and it has excellent prompt adherence.
That it does furries real well is just the cherry on top of it all.
dude just discovered the furry fandom.
Well... since the Egyptian...
It works quite well as a general model.
Malebolegia generates a LOT of his image samples for his Lora on this model. This is not just for furry.
@LuminalArt Absolutely. You can blame 90s cartoons, cereal mascots, Zootopia, Nintendo, Beastars anime (VERY good story on Netflix) and more... Sci-fi and fantasy role play span quite wide! :-)
@LuminalArt Fur is just more attractive than skin. Beauty is in the eye of the beholder.
Hi! Really like your models.
Curious if you will be expanding to Z-image series now that the base model is available?
Oh damn new version.
Rubs hand together
Lets see how it is
Edit: I don't see much improvement over previous versions. Comparing to V14 which I normally use it just does more silly errors compared to it like giving two belly buttons or adding extra colors to a character that I just want to have one color.
I can't seem to be able to select nova furry 16
Wait until Friday 0:00AM GMT
Or you can use it via other site such as PixAI or SeaArtAI via description's link
A little heads up. It seems that the IL v16.0 model has an identical AUTOV2 hash value with this other CivitAI resource (https://civitai.com/models/1934382?modelVersionId=2474301), so do keep it in mind when uploading images from external generators and collecting metadata automatically. CivitAI may assign this other resource as the one used instead of Nova Furry XL IL v16.0, in which case you need to add the image metadata manually. If you're not careful, you might accidentally post your image under the wrong model just like I almost did.
UPDATE: It has been fixed now.
I can't seem to get eyes look as good as they do in the example gallery, when I try it I often have pupils and iris mix together or not well defined, and definitely never as clearly defined as here. Does anyone have some advice?
Try hires.fix with x1.5 resolution, 0.4 denoise, 16 steps and 4.5 CFG
For myself, I use a eye detailer lora on the 1024 iterations, 35 steps, then hires.fix up to 2x or 1.5x, 0.3 denoise, 15 high res steps. That seems to do very well, but yeah, the 1024 base iterations have much less clear eyes.
I genuinely do not understand how this model is so good. It just produces banger after banger.
Anyone seeing the default posing or dynamic pose leaning heavily to selfies if you try to do a front view of the character?
Somewhat... I have pretty good luck with "front view, head on view". What I do often get is a double character - someone pocketed inside another's shirt lol
I've not seen that, but it's all in the prompt and the aspect ratio of the image. Some tags are going to lean selfie, and so are some aspect ratios.
Absolutely. I put in the negative prompt "(reaching_towards_viewer:1.80)" and it stopped doing it.
@sgtp3 That double character, often tiny or strange, is something I see crop up occasionally when the tag "head on..." are included, even if it's "head on view". There are several odd ones like that.
@RykOakwine Makes sense. Sucks too, that was a useful token in past versions. Petition for the creator to rename it to cleavage pet or something and leave head on alone lol
it's perfect bro
Sometimes when I ask it to generate a complete body, it generates one with a top-down perspective and it looks a bit strange. How can I prevent it from generating with that perspective?
Can you try "full body" or "full scene"?
@Soosage Yeah, and I also tried changing the “cfg” and the steps, adjusting the resolution, etc. It creates a top-down perspective and makes the legs look weird.
I use "(full-length_portrait:1.2)" to get the entire body.
For the view angle, what do you currently use in your prompt? Try to avoid using anything with "dynamic" as that will tell the AI to figure out an angle or pose based on the context.
Otherwise, try "front_view" or maybe "three-quarter_view, three_quarter_angle" for a bit of a side view.
@dragonofdarkness1992 Thanks, I'll give it a try
I know this doesn’t have anything to do with the perspective but If you mean the body looks warped or short then maybe putting “foreshortening” in the negative prompt could help avoid that?
@sniping643 That's not it—the thing is, sometimes I want to generate a full-body image of a character, but it ends up looking like a selfie taken from above, and while that looks okay sometimes, it often makes their legs look distorted, haha.
I've been using this model quite a bit, but I'm having problems with associated tags on the newer versions.
When I want to generate my shark girl just from the tags, I always get back/dorsal fin, especially when adding "view from behind", despite me adding "back fin" and "dorsal fin" to negative prompt with increased weights. Old versions were able to work with it and rarely generated the fin.
I'm using Euler a with SGM Uniform with 50 steps and CFG set to 5.0.
Also, I want my generations to remain private, at least for now.
And don't get me wrong, this is my favorite model despite this problem. Just wanting to share my "bug report" and maybe receive some support on how to manage this issue better on my end.
Thanks!
I've found NegPiP extension that helped me with this problem.
Adding "(dorsal_fin:-1.4), (back_fin:-1.4)" to the prompt removed the fin I didn't want.
By the way, if anyone wants to try it, just be cautious with the negative tags - this plugin sometimes works too well. When I added "shark fin", my shark stopped being a shark.
Hey, you have any suggestion how to move the caracter(s) from the dead center? I have already tried prompting /with a controlnet in Comfy, but I have no luck with it...
regional prompter
@platypusFTW Thanks for the suggestion, I'll take a look into it.
@Crovidius7849 Forge couple works pretty well (if your using Forge Neo)
@Asaghon (I see, I'll give it a shot if that works on Comfy desktop) Although right now I try to make it work an openpose controlnet...
*I used to use Forge, but I felt system limitations so I switched to Comfy (after my comment saw what exactly you suggested, sry for the confusion on my end).
@Crovidius7849 I think there is Forge couple for Comfy as well, but haven't tried it. Can't see any limitations for anime content tbh. Got plenty of images with 2/3 distinct characters using just Forge Couple in Forge Neo without any extra lora's.